diff --git a/docs/configuration.mdx b/docs/configuration.mdx
index b2ada9c..4ad2061 100644
--- a/docs/configuration.mdx
+++ b/docs/configuration.mdx
@@ -152,7 +152,7 @@ Supported actions:
 1. block
 2. log
 
-Unlike PII Redactor, `entities` defaults to only the high-risk set: `credit_card`, `api_key` and `bearer_token`. Contact data and locators are supported but opt-in. In a proposed tool argument, they are usually the tool's own parameters rather than an exfiltration signal. `scan_injection` is off by default for the same reason. See the [Tool Approval Guard guide](/middleware/tool-approval-guard#secret-like-data).
+Unlike PII Redactor, `entities` defaults to only the high-risk set: `credit_card`, `api_key`, and `bearer_token`. Contact data and locators are supported but opt-in. In a proposed tool argument, they are usually the tool's own parameters rather than an exfiltration signal. `scan_injection` is off by default for the same reason. See the [Tool Approval Guard guide](/middleware/tool-approval-guard#secret-like-data).
 
 `block_entities` stops the run regardless of the `action` setting. For an observe-only `action: 'log'` setup, also set `block_entities` to `[]`.
 
diff --git a/docs/middleware/budget-guard.mdx b/docs/middleware/budget-guard.mdx
index 5a096f9..22c34e6 100644
--- a/docs/middleware/budget-guard.mdx
+++ b/docs/middleware/budget-guard.mdx
@@ -133,7 +133,7 @@ Once the provider replies, the middleware prices the spend recorded against the
 
 ### Pricing
 
-A built-in table prices common models from Anthropic, OpenAI, Google, xAI, DeepSeek, and Mistral, in US dollars per million tokens. The middleware matches a model ID exactly first, then by the longest pricing key it starts with or contains. A dated or gateway-prefixed id such as `us.anthropic.claude-3-5-haiku-20241022-v1:0` or `bedrock/gpt-4o-mini-2024-07-18` still resolves to its family.
+A built-in table prices common models from Anthropic, OpenAI, Google, xAI, DeepSeek, and Mistral, in US dollars per million tokens. The middleware matches a model ID exactly first, then by the longest pricing key it starts with or contains. A dated or gateway-prefixed ID such as `us.anthropic.claude-3-5-haiku-20241022-v1:0` or `bedrock/gpt-4o-mini-2024-07-18` still resolves to its family.
 
 Published pricing changes faster than this package releases. Override or extend it:
 
@@ -148,8 +148,8 @@ new BudgetGuard(
 
 <Warning>
     When no pricing is known for a step's model, `unknownModelAction` decides what happens:
-    `allow` (the default) logs the gap and lets the request through unmetered, while `block`
-    stops the run so an unpriced model can never spend against a budget unchecked.
+    `allow` (the default) logs the gap and lets the request through unmetered. `block` stops
+    the run so an unpriced model can never spend against a budget unchecked.
 
     ```php
     new BudgetGuard(unknownModelAction: 'block')
@@ -170,8 +170,8 @@ new BudgetGuard(action: 'log') // observe before enforcing
 
 <Warning>
     `truncate` only applies to the per-request budget. Shrinking the current prompt cannot undo
-    spend already recorded against a session or a user's day, so a session or daily breach
-    degrades to `block` regardless of the configured action.
+    spend already recorded against a session or a user's day. A session or daily breach
+    therefore degrades to `block` regardless of the configured action.
 </Warning>
 
 ## Blocked runs
diff --git a/docs/middleware/tool-approval-guard.mdx b/docs/middleware/tool-approval-guard.mdx
index 79d408b..ae61e7e 100644
--- a/docs/middleware/tool-approval-guard.mdx
+++ b/docs/middleware/tool-approval-guard.mdx
@@ -90,7 +90,7 @@ new ToolApprovalGuard(
 
 ### Secret-like data
 
-The middleware runs the same detectors as [PII Redactor](/middleware/pii-redactor), including the Luhn check on card numbers. By default it scans for **only three entities**: `credit_card`, `api_key` and `bearer_token`.
+The middleware runs the same detectors as [PII Redactor](/middleware/pii-redactor), including the Luhn check on card numbers. By default it scans for **only three entities**: `credit_card`, `api_key`, and `bearer_token`.
 
 Entities listed in `block_entities` stop the run regardless of the configured action, matching PII Redactor's behaviour.
 
@@ -102,8 +102,8 @@ Entities listed in `block_entities` stop the run regardless of the configured ac
 
     Exfiltration is about **destination**, not presence, and the destination
     cannot be judged from the value alone. Adding `email`, `phone`, `url`,
-    `ip_address` or `mac_address` will flag legitimate calls made by any agent
-    that owns a mail, SMS or HTTP tool.
+    `ip_address`, or `mac_address` will flag legitimate calls made by any agent
+    that owns a mail, SMS, or HTTP tool.
 
     To control *where* a tool may send data, use `allowed_tools` /
     `denied_tools` and domain allowlisting inside the tool itself.